05 International Diversification With ADRs
-
Upload
abdul-azim-zainal-abidin -
Category
Documents
-
view
224 -
download
0
Transcript of 05 International Diversification With ADRs
-
8/8/2019 05 International Diversification With ADRs
1/34
INTERNATIONAL DIVERSIFICATION WITH
AMERICAN DEPOSITORY RECEIPTS (ADRs)
By:
M. Humayun KabirSenior Lecturer in Finance
Department of Finance, Banking & PropertyMassey University
Palmerston North, New Zealand
and
Neal MaroneyAssociate Professor
Department of Economics & FinanceUniversity of New Orleans
and
M. Kabir Hassan
ProfessorDepartment of Economics & FinanceUniversity of New Orleans
Visiting ProfessorDrexel Univesity
-
8/8/2019 05 International Diversification With ADRs
2/34
Abstract
It is already well known that U.S. investors can achieve higher gains by investing directly in emergingmarkets (De Santis, 1997). Given the opportunity to invest directly in the shares of stocks in the developed(DCs) and emerging (EM) markets, it is interesting to know whether the U.S. investors can potentially gainany benefits by investing in ADRs. We test both index models, and SDF-based model.Our findings showthat U.S. investors needed to invest in both ADRs and country portfolios in developed in the eighties, andin Latin American countries in early nineties. During the early and late nineties, we find substitutabilitybetween ADRs and country portfolios in DCs. As more and more ADRs are enlisted in the US market fromdeveloped countries over time, the ADRs become substitutes to country. Similarly, countries with highernumber of ADRs irrespective of regions show the same pattern of substitutability between ADRs andcountry indices. However, such substitutability does not exist for countries with the highest number of ADRs by the end of sample period, 2001. On the other hand, U.S. investors can achieve the diversificationbenefits by investing ADRs along with U.S. market index in Asia. The significant marginal contribution of one-third of developed countries requires investment in ADRs and U.S. market in the developed countries.And investors do not need to hold both ADRs and country as it was the case in the eighties. On the otherhand, investors need to hold both ADRs and country portfolios in most of the Asian countries to achieve
diversification benefits at margin.
-
8/8/2019 05 International Diversification With ADRs
3/34
1. INTRODUCTION
With the globalization of capital markets, an increasing number of foreign firms
have chosen to enter the U.S. market with the issuance of American Depository Receipts 1
(ADRs) in order to broaden the shareholders base, raise additional equity capital by
taking advantage of liquidity of U.S. market. Over the last decades the number of foreign
firms listed as ADRs in the U.S. market has gone up dramatically. According to Bank of
New York, by the year 2000, the number of ADRs have risen to about 2,400, of which
about 600 are traded on NYSE, AMEX or NASDAQ, and the remaining on the OTC 2.
Lins, Strickland, and Zenner (2000), in their recent study, found that the greater access to
external capital markets is an important benefit of a U.S. stock market listing, especially
for emerging markets firms. The enlisted foreign firms that are subject to SEC reporting
and disclosure requirements, reduce informational disadvantages, and agency costs of
controlling shareholders due to better protections for firms coming from countries with
poor investors rights. As a result, firms with higher growth opportunities coming from
countries with poor investors rights are valued highly (Doidge, Karolyi, and Stulz,
2001).
The objective of the present study is to find the diversification potential of ADRsin different regions, and in countries from the perspective of an U.S. investor. Given the
1 The ADRs are negotiable certificates or financial instruments issued by U.S. depository banks that holdthe underlying securities in the country of origin through the custodian banks The ADRs denominated in
-
8/8/2019 05 International Diversification With ADRs
4/34
opportunity to invest directly in the shares of stocks in the developed (DCs) and emerging
(EM) markets, it is interesting to know whether the U.S. investors can potentially gain
any benefits by investing in ADRs. It is already well known that U.S. investors can
achieve higher gains by investing directly in emerging markets (De Santis, 1997). Since
ADRs are traded in the U.S. market they have been considered as an alternative to such
cross-border investments while ensuring a higher diversification benefits. However, a
small fraction of these ADRs are, in fact, enlisted on major U.S. exchanges. Most of the
ADRs are unlisted, and traded OTC (level I), and they maintain home country accounting
standard, and do not require SEC registration. Only level II and level III ADRs are
enlisted, and comply with SEC regulations 3. These ADRs can be considered as the subset
of country shares. Solnik (1991) argues that the ADRs traded in the U.S. market are
mostly big firms in their home countries, and it is likely that they have lower
diversification benefits than a typical foreign firm. As s result we pose the question: can
ADRs provide as much diversification gains as the country indices?
Most of the studies on ADRs (Hoffmeister, 1988; Johnson and Walter, 1992;
Wahab and Khandwala, 1993; Callaghan, Kleiman, and Sahu, 1996; Jorion and Miller,
1997) show how combining ADRs with U.S. market or other funds can reduce risk
without sacrificing expected returns. A host of studies (Parto, 2000; Choi and Kim, 2000;
Kim, Szakmary and Mathur, 2000; Alanagar and Bhar, 2001) find the determinants of
ADR returns in the context of a single or multi-factor models. Bekaert and Uris (1999)
-
8/8/2019 05 International Diversification With ADRs
5/34
contribution of the present study is that we attempt to show how the case of
diversification with ADRs varies not only across regions over different sample periods
but also with the size of ADR markets measured by the number of ADRs irrespective of
country of origins. We also explore the possible combination of assets the U.S. investors
require to hold in different regions and countries with respect to diversification. We
address the shortcomings of an ill-defined benchmark conduct in standard empirical tests
like index models. If the market portfolio is not mean variance efficient, one can
incorrectly conclude real assets are good diversifiers. Secondly, the case for
diversification depends on the temporal stability and significance of a set of assets in an
investors portfolio, and these assets correlation structure. If asset correlation vary
widely, then diversification benefits are questionable as an optimized portfolio becomes
expensive or impossible to maintain in the face of uncertain correlations. Unlike the
index models, spanning tests are not subject to benchmarking error, as they do not rely on
a specific benchmark asset pricing model. 4 We use spanning test proposed by Hansen
and Jagannathan (1991) based on stochastic discount factors (SDFs).
2. A REVIEW OF RELEVENT LITERATURE
There are some studies that concentrate on ADRs returns behavior, their
determinants, and the opportunities for diversification gains in both U.S. and international
context. Officer and Hoffmeister (1988) show that ADRs lower portfolio risk when added
to portfolio of U.S. stocks. In fact adding as few as four ADRs in a representative U.S.
-
8/8/2019 05 International Diversification With ADRs
6/34
underlying shares of developed countries mostly traded on NYSE and AMEX exchanges
for the period of 1973-1983.
Wahab and Khandwala (1993) use weekly and daily return data of 31 pairs of
ADRs and the underlying shares of mostly developed countries (UK, Japan, France,
Germany, Australia, S. Africa, Sweden, Norway, Luxembourg), and with the use of
active portfolio management strategies they show that ADRs provide expected returns
that are similar to their respective underlying shares. However, the greater the investment
proportion in ADRs, or the higher the number of ADRs given assumed investment
weights, the larger the percentage decline in daily returns on the combined portfolio in
comparison to the standard deviation of returns on S&P500.Thus ADRs potentially
provide better risk reduction benefit, and have been stable over several sub-periods.
Johnson and Walther (1992) show that while combining the ADRs, direct foreign
shares, and international mutual funds substantially increases portfolio returns per unit of
risk, ADRs and direct foreign shares alike would have offered more attractive portfolio
risk and return improvements when compared to domestic diversification strategy.
Callaghan, Kleiman, and Sahu (1996), using Compustat data of 134 cross-listed
firm for 1983-1992 sample period, find that ADRs have lower P/E multiples, higher
dividend yields, lower market-to-book ratios than international benchmark, as measured
by MSCIP. Moreover, ADRs provide a higher monthly return and a higher standard
deviation than the MSCIP, while both the ADR sample and MSCIP have lower betas than
-
8/8/2019 05 International Diversification With ADRs
7/34
index. And ADRs receipts can be used to replicate or track the emerging market IFC
index.
Alaganar and Bhar (2001) find that ADRs have significantly higher reward-to-risk
than underlying stocks. On the other hand, ADRs have a low correlation with the US
market under high states of global and regional shocks. Kim, Szakmary, and Mathur
(2000) document that while the price of underlying shares are most important, the
exchange rate and the US market also have an impact on ADR prices in explaining ADR
returns. Parto, 2000; Choi and Kim, 2000 find that local factors (market and industry),
and their underlying stock returns across both countries and industries better than the
world factor, especially for emerging markets. On the other hand, a multi-factor model
with world market return and the home market return as the risk factors performs better
than models with just the world return, the home market return or a set of global factors
as the risk factors.
Bekaert and Uris (1999) conduct a mean-variance spanning test for closed-end
funds, open-end fund and ADRs of emerging market with some global return indices
such as FT-Actuaries, U.S. index, U.K. index. European less U.K. index, and Pacific
index as the benchmark. For comparison, they also examine the diversification benefits of
investing in the corresponding IFC investable indexes. The sample period consists of two
sub-periods: September 1990 August 1993 for closed-end funds, and September 1993
August 1996 for closed-end, opened-end funds, and ADRs. The study finds that the U.S.
-
8/8/2019 05 International Diversification With ADRs
8/34
3. HYPOTHESE AND METHODOLOGY
3.1 Motivation and Hypotheses
De Santis (1997) has found that an U.S. investor can gain diversification benefits
by investing directly in Emerging markets, not in developed markets. He found higher
reward-to-risk performance from investing in EM countries. However, Latin American
countries provide the highest level of diversification countries. With recent popularity of
ADRs as an investment vehicle raise a reasonable question: Is ADR an alternative to
investing directly in stocks in different countries? In other words, can a U.S. investor
achieve diversification benefits beyond what is achievable through investing directly in
country index? Solnik (1991) expresses his doubt on such diversification benefits. The
ADRs traded in the U.S. market are mostly big firms in their home countries, it is likely
that they have lower diversification benefits than a typical foreign firms. As a result, the
efficient frontier is expected to shift in mean-variance space when the test asset, i.,e., the
portfolio of ADRs is excluded if there really exits any gains for ADRs. Secondly, the
level of economic integration and asymmetry of information are different between
developed and emerging countries, even within the emerging countries. As a result, we
want to test such diversification benefits across countries, and across regions (developed,
Latin America, Asia, Emerging markets). Moreover, a huge number of ADRs, especially
from emerging countries, have been enlisted in the U.S. market in 1990s. It is also
interesting to look how such influx of foreign shares in the U.S. market has changed the
-
8/8/2019 05 International Diversification With ADRs
9/34
implies an achievable diversification benefits by investing in ADRs, and U.S. investors
can gain, at least, the same level of benefits from their investment in ADRs, which can be
regarded as the substitute of cross-border investment. However, substitutability may not
be the only outcome. Investors may require to hold both ADRs and country indices. As a
result, we further ask: Can country and U.S. market indices mimic the ADR returns? The
following table the implications of our spanning questions:
Can country and U.S. market mimic ADR return?
YES NO
YES ADRs and country aresubstitute s for each other
ADR required,country not required
Can ADRs and U.S.market indicesmimic countryreturns? NO Country portfolios required,
ADR not requiredBoth ADR and countryrequired
If country indices and U.S. market can mimic ADR returns, then ADRs become
redundant. On the other hand, if ADR portfolios and U.S. market can mimic country
returns, investors do not need to hold country indices. The latter scenario is most unlikely
as the ADRs are the sunsets of country stocks. In case country and U.S. market mimic
ADR returns, but ADR and market cannot mimic country returns (NO-YES in the above
table), no achievable diversification benefits to the U.S. investors as proposed by Solnik.
However, in NO-NO situation, investors need to hold both country and ADR portfolios to
achieve the benefits spanned by the country, ADR, and U.S. market.
-
8/8/2019 05 International Diversification With ADRs
10/34
We use both index models, and spanning test proposed by Hansen and
Jagannathan (1991), De Santis (1993), Bekaert and Uris (1996), and Maroney and
Protopapadakis (1999) based on stochastic Discount factors (SDFs).
3.2.1 Index models
The index model with risk-free rate is defined as
, , , , ,c i i a i a i m i m i ir r r = + + + 1,.....,i N = (1)
where, r c,i , r a,i , and r m,i are the excess returns over risk free rate of country, ADR, and
U.S. market portfolio returns for each country. The null hypothesis is:
0 H : 0i = i (2)
The rejection of null hypothesis implies ADR portfolios are not enough to achieve as
much diversification benefits as the country indices can provide. In other words, when
the null is not rejected, a linear combination of ADR returns and U.S. market returns can
replicate the returns of country indices.
We apply Seemingly Unrelated Regression (SUR) estimation technique for the
system of equations in (1). And in order to test the restriction in (2), we use Wald test,
which reports chi-square statistics with degrees of freedom equal to the number of
restrictions in (2). We also reports the likelihood ratio test, which reports the chi-square
statistics with degrees of freedom equal to q, T-k, where q is the number of restrictions, T
is the number of observations, and k is the number of parameter to be estimated. Both
W ld d LR i ll i l b diff
-
8/8/2019 05 International Diversification With ADRs
11/34
The test reveals the necessary condition for the efficiency of a linear combination of L
returns with respect to the total set of N+L risky assets given in the null in (2). The
noncentral F distribution with degrees of freedom N and (T-N-L) is given as:
( )1
1 10 0
11 p p
T T N Lr r
N T L
+
(3)
where pr is a vector of sample means for ( )1 2, ,......., pt t t Lt r r r r = , is the sample
variance-covariance matrix for pt r , 0 has a typical element 0i , and 0 is the least
squares estimators for 0 based on the N regressions in (1). The noncentrality parameter
is given by
( )1 10 0 / 1 p pT r r + (4)
Under the null hypothesis in (2) 0 = , and we have a central F distribution.
In absence of any common risk-free rate, which is plausible in real world since
risk-free rates differ across countries, investors may choose risky portfolios zero-beta
portfolios - from the set of efficient frontier portfolios. In such case, benchmark
portfolios include one more asset. In order to test whether benchmark portfolios matter,
we test the following hypothesis:
0 H : i = i (5)
In other words, we test whether the alphas are same across countries. We use Wald and
Likelihood ratio tests for that purpose
-
8/8/2019 05 International Diversification With ADRs
12/34
benchmark assets common to both frontiers and the other q test assets are only included
in the construction of one of the frontiers. The frontier MV p+q will always encompass
MV p because it contains more assets used in its construction. As mechanics dictate, the
Sharpe ratio Sp+q is necessarily greater than Sp. The null hypothesis of spanning states
that both frontiers statistically coincide,
H0: MV p+q = MV p . (6)
It is sufficient to measure the distance between frontiers at two points, as all other points
are convex combinations 5. A confirmation of the spanning hypothesis implies that
additional test assets q will not offer diversification opportunities relative to those already
included in the portfolio of benchmark assets p. In other words, the set of benchmark
assets prices the broader set of assets. Evidence against the spanning hypothesis means
the inclusion of test assets takes advantage of diversification opportunities not available
in the benchmark assets, thus these test assets should be included in any well-diversified
portfolio.
Spanning-like tests can be conducted with a variety of asset pricing models. For
example, the Capital Asset Pricing Model restriction that constant term in the CAPM
regression of excess returns be zero is the less restrictive intersection hypothesis that tests
the distance at one point (test to see if Sp+q = Sp ). This approach is problematic as this is
a joint test of the specific discount factor created by the CAPM and the intersection
hypothesis. The claim that the market portfolio is on the efficient frontier of all assets is
-
8/8/2019 05 International Diversification With ADRs
13/34
approach to spanning tests is that it guarantees portfolios used in spanning tests lie on the
mean variance frontier of assets used in their construction. These tests are based on
Hansen and Jagannathan (1991), HJ, volatility bounds that give a lower boundary on the
volatility of all stochastic discount factors SDFs such that the Law of One Price (LOP) is
not violated. HJ bounds have a dual relationship to the mean variance frontier because
SDFs are portfolios constructed to have the lowest variance possible for a given return.
Spanning tests exploit the mean variance efficiency of the HJ bounds: selecting two
points on the mean variance frontier is the same as selecting two points on the HJ bounds.
HJ bounds are model free in the sense they do not require one to specify a specific
benchmark model and therefore avoid the joint hypothesis problem inherent in using
CAPM or Multifactor benchmarks. HJ (1991) use the LOP to derive the bound on
admissible discount factors. The LOP is the minimum restriction on asset prices that
assets which have the same set of payoffs sell for the same price. The LOP is the present
value relation at the heart of asset pricing:
( ) t t t Pm X E =++ 11 , or equivalently ( ) 111 =++ t t m R E , (7a, 7b)
where X , and R are a t+1 payoffs and gross returns on N assets, m is the SDF and P is
vector of todays asset prices. The SDF m is commonly called the intertemporal marginal
rate of substitution, which places additional restrictions on its construction. If there is no
such restriction, it is straightforward to derive the SDF m as an algebraic exercise.
-
8/8/2019 05 International Diversification With ADRs
14/34
1],[][][][ =+= m RCovm E R E Rm E . (8)
HJ prove m is linear in returns:
[ ] 'r m E m += , (9)
where r=R-E[R] is vector of N return deviations from their times series means, is a
vector of weights on N assets, and E[m] is the expected value of the discount factor.
Now solve for the set of weights such that the solution satisfies (8). Substituting (9) into
(8) and simplifying gives,
=+ ][][][ r r E m E R E . (10)
Solving for we see,
( )][][1 m E R E = , (11)
where is the covariance matrix of returns. The discount factor has variance,
[ ] =mVar (12)
which has the lowest variance of any candidate discount factor because the SDF isconstructed to satisfy (7b) with a linear projection on the payoffs (refer to Campbell, Lo
and MacKinlay (1997), or Cochrane (2001) for proofs of propositions related to the
6
-
8/8/2019 05 International Diversification With ADRs
15/34
E[m] as its expectation can be constructed solely from returns data. The SDF will have
the lowest variance of any candidate SDF with the same expectation. The expectation of E[m] has the property of being the price of the risk free asset.
To form the parabola that is the HJ volatility bound, HJ treat E[m] as an unknown
parameter c, and the standard deviation of the SDF with expectation c, (mc) is plotted
against c. This forms the lower bound on all SDFs constructed from a set of returns that
satisfy the LOP. Any other SDF must have higher variance than the one along the bound
to price all the assets.
Spanning tests use two SDFs with expectations E (m1), and E (m2) chosen in a
reasonable range to measure the difference between frontier portfolios at two points.
Following Maroney and Protopapadakis (2002), the empirical form of spanning tests is:
( ) ( )
( ) ( ) ,0,,0,
222
111
==
==
E m R E R
E m R E R(13)
where,
( )( )
( ) j j
j m E
Rmm R E
cov1 = , and ( ) qjq pj p j j r r m E m ++= ; { }2,1= j .
With N assets there are 2N orthogonality conditions and without restrictions the system is
just identified and linear with coefficients { }2211 ,,, q pq p . Without restrictions all
orthoga1ity conditions are satisfied and sample averages are replicated using either SDF--
by construction. The restriction given by spanning implies the two SDFs produced from
-
8/8/2019 05 International Diversification With ADRs
16/34
need to include test assets in the construction of the SDFs: { }021 == qq .
Overidentifying conditions reveal how well the SDF produced from p assets replicates
the sample averages of the broader set. We estimate the system using GMM with a
Newly and West (1987) correction for first order autocorrelation. The Hansen (1982) J -
Statistic based on the criterion from GMM and distributed ( )q22 will evaluate the
goodness of fit of the overidentifying conditions.
3.2 Data
Our sample consists of NYSE/AMEX/NASDAQ-enlisted ADRs of 27 countries
from developed, Asian, Latin American market. Table 2 presents the number of ADRs
for each country by the year they have been introduced in the U.S. market. We selected
only the ADRs for which the data are available in CRSP. We observe a proliferation of
ADRs with more countries in the 1990s. We construct monthly country-level portfolios
of returns (equally-weighted) of those ADR programs from CRSP database for the period
1981 2001. Since developed countries entered U.S. market earlier, and the Lain
American or Emerging markets are late starter, a long time span that includes most recent
periods is an essential element to capture the implications for return dynamics. We divide
the sample period into three sub-periods: 1981-1990, 1991-1995, and 1996-2001. All
country indices are collected from MSCI. We use U.S. market index of MSCI as the
benchmark of U.S. investors.
4 RESULTS ANALYSIS
-
8/8/2019 05 International Diversification With ADRs
17/34
different countries during different periods. Without exception, ADR portfolios have
higher variances uniformly. The t-statistics are used to compare the mean returns betweenthe country and ADR indices for each country. We find significantly different mean
returns for half of countries in the early 1990s, and for one-third of countries during late
1990s. In Asian region mean returns are statistically different for five out of six countries.
Table 4 shows the own correlation between ADR portfolios and country indices for
different countries in three sub-periods. We find, in general, correlation are less than
perfect creating an opportunity for benefits form investments in ADRs, and an increase in
correlation subsequent periods for a majority of countries after the introduction of ADRs
in U.S. market.
4.2. Index Model Results
In Table 5, we presented the probability of Wald, LR, and GRS tests. While the
probability of rejections of Wald test are based on chi-square distribution, the probability
of rejections of LR and GRS tests are based on F-distributions. During the first sub-
period, 19981-1990, our same includes only countries from developed regions. For
developed countries the hypothesis of alphas equal to zero jointly is rejected at a very
high level of significance. We find similar results during the early nineties. This signifies
that a linear combination of U.S. market returns and ADR returns of developed countries
cannot span the countries national returns. In other words, an U.S. investor can not
achieve the same level of diversification benefits by investing ADRs with U.S. market
-
8/8/2019 05 International Diversification With ADRs
18/34
reject the hypothesis for developed regions. Similarly, we reject the null according to
GRS test only for Latin America. In Asia, all test tests report significantly different fromzero alphas jointly.
We report the test results of the null hypothesis that the intercepts are joint same
across portfolios. Except for developed regions, the hypothesis is rejected for all
countries in Latin America and Asia. This is plausible as the developed countries are
more integrated economically, whereas the emerging markets face varying levels of
integration so that zero-beta returns vary across countries. This finding also indicates that
the benchmark assets matter in pricing the emerging markets assets, and risk-free rates
are not same across countries.
4.3 Spanning Test Results
4.3.1 Regional diversification benefits
In case of regional diversification test, we our benchmark assets are all country
indices, all ADR portfolios and the U.S. market except the set of country indices (ADR
portfolios) for a region, which are considered as test assets. The spanning test results for
three regions are presented in Table 6. For the set of countries, we find that the
overidentifying conditions are rejected irrespective of whether we choose country indices
or ADR portfolios as test assets during different sub-periods except for DCs for periods
19991-1995, and 1996-2001, and for Asian region when country indices are used as test
assets. In cases where null hypotheses for both (country indices as test assets or ADR
-
8/8/2019 05 International Diversification With ADRs
19/34
As a result, U.S. investors need to invest both in ADRs and country portfolios. In other
words, we find the diversification potentials in the developed region during the earlyperiods, 1981-1990, and in Latin American region. For the early and late nineties, we
cannot reject the hypothesis at 5 percent level implying that both ADRs and country
portfolios can be substituted for each other in DCs. During the period 1981-1990, we
have only nine developed countries, which have ADRs in the U.S. market. The total
number of ADRs during this period was 90 as shown in table 2. We observe an explosion
of ADRs in later periods form developed countries, and more countries in the U.S. ADR
market. Such proliferation of ADRs may have changed the market. An increase in own
correlation between ADR returns and country indices indicate a more integrated nature of
market. As a result, investors can substitute the ADRs for country with no additional
benefits to gain. We find independent diversification benefits in Latin American
countries. Only for Asian countries, we cannot reject the null when country indices are
used as test assets, but it is rejected when we use ADR portfolios as test assets implying
that U.S. investors need to invest in ADR portfolios with market index to achieve
diversification benefits.
Since our sample sizes are small; and spanning tests can be potentially bias with
such small sample. In order to check that we use the bootstrap technique, for which we
use 3,000 iterations by randomizing the e i= r - E(r) , and generate J-statistics sample.
Based on the sample we find the probability of our original J-statistics from the GMM
-
8/8/2019 05 International Diversification With ADRs
20/34
4.3.2 Portfolios sorted by number of ADRs
We observe a significant increase in ADRs in the 1990s from different countries.We also find substitutability between ADR portfolios and country indices in the later
periods for developed region, whereas they provide diversification benefits independently
in the 1980s. In order to find whether such proliferation of ADRs each country became
representative of those countries so that investors do not need ADRs in their portfolios,
we form portfolios based on the number of ADRs irrespective of regions. For each t-
period returns, we form equally-weighted portfolios based on ( t-1)- period ranking of
countries with respect to the number of ADRs for each country. The reason to rank
countries is to divide the countries in three different groups: largest, medium, and
smallest. We continue to follow the same process for each years starting form 1980 to
2001. Then we stacked the data for the whole period, and conduct our spanning test. The
benchmark assets consist of all three largest, medium, and smallest - ADR portfolios,
three equally-weighted country portfolios, and the U.S. market except one of the ADR
(country) portfolios that is considered as test asset.
The spanning test results are presented in Table 7, panel A. We find that the
overidentifying conditions are rejected for all sub-periods for all portfolios whether we
choose country or ADR portfolios as the test assets except the largest one in the third sub-
period. In other words, the null hypothesis that the ADR and the U.S. market can mimic
country or the country and the U.S. market can mimic cannot be rejected at conventional
-
8/8/2019 05 International Diversification With ADRs
21/34
the nulls are rejected showing the requirement to hold both ADRs and country in the
portfolios.Unlike forming equally-weighted portfolios as previous, we also divide country
and ADR portflios into the largest, medium, and the smallest groups based on the number
of ADRs in the year 1990, 1995, and 2001 irrespective of regions. The spanning test
results are presented in panel B of table 7. We find that the largest portfolio in the third
sub-period, 1996-2001, cannot reject the null that the country plus the U.S. market index
can mimic the ADR return, but it rejects the null that ADRs plus the U.S. market can
mimic the country return. This implies that ADRs cannot provide diversification benefits
for countries that have the highest number of ADRs by the end of sample period. For all
other period, we the null hypotheses are rejected implying that the investors need to hold
both ADRs and country to gain benefits in the market.
4.3.3 Marginal diversification benefits
We attempt to evaluate the contribution of each country at margin towards
diversification benefits with respect to the grand mean-variance frontier composed of all
ADR and country portfolios, and the U.S. market index. Each time we exclude one
country ADR portfolio (one country index), find the probability of rejection of the null
hypothesis that returns of all countries portfolios, ADRs portfolios and U.S. market can
mimic that countrys ADR returns (returns of ADRs portfolios, all other countries
portfolios and U.S. market can mimic that countrys returns) from corresponding J-
-
8/8/2019 05 International Diversification With ADRs
22/34
diversification benefits for each country are presented in table 8 and 9. We find that in the
early period, Denmark, and Ireland show substitutability between ADR and country. Formajority of the developed countries, U.S. investors need either both (No-No) of them or
country (No-Yes) portfolios with U.S. market index. During the period, 1991-1995, most
of the countries get transition from No-No situation to Yes-Yes situation implying
substitutability between ADRs and country. During the period 1996-2001, we explore the
similar trend with some countries with no diversification benefits, or requiring
investments in ADRs. We find, similarly, an increase in substitutability in Latin America
for the period 1996-2002. However, in Asia, for majority of countries, investors need
both ADRs and country to achieve benefits from investment.
5. CONCLUSION
The objective of the present study is to measure the diversification benefits of
different country ADRs portfolios from the perspective of an U.S. investor. Studies on
ADRs show indirect evidence of achievable diversification benefits. Given the
opportunity to invest directly in the shares of stocks in the developed (DCs) and emerging
(EM) markets, it is interesting to know whether the U.S. investors can potentially gain
any benefits by investing in ADRs. Our findings show that U.S. investors needed to
invest in both ADRs and country portfolios in developed in the eighties, and in Latin
American countries in early nineties. During the early and late nineties, we find
substitutability between ADRs and country portfolios in DCs. As more and more ADRs
-
8/8/2019 05 International Diversification With ADRs
23/34
However, such substitutability does not exist for countries with the highest number of
ADRs by the end of sample period, 2001. On the other hand, U.S. investors can achieve
the diversification benefits by investing ADRs along with U.S. market index in Asia. The
significant marginal contribution of one-third of developed countries requires investment
in ADRs and U.S. market in the developed countries. And investors do not need to hold
both ADRs and country as it was the case in the eighties. On the other hand, investors
need to hold both ADRs and country portfolios in most of the Asian countries to achieve
diversification benefits at margin.
-
8/8/2019 05 International Diversification With ADRs
24/34
References
1. Bekaert, G, 1996, Market Integration and Investment Barriers in Emerging EquityMarkets.
2. Bekaert, G, and M. S. Urias, 1999 Is there a free lunch in emerging marketequities?, Journal of Portfolio Management , Vol 25, No. 3, 83-95.
3. Callaghan, J.H., R. T. Kleiman, and A. P. Sahu, 1996 The investment characteristicsof American depository receipts, Multinational Business Review, Vol. 4, 29-39.
4. De Santis, G., Volatility Bounds for SDFs: Tests and Implications from InternationalStock Returns, 1993, Unpublished Ph.D. dissertation , University of Chicago.5. De Santis, G., Asset pricing and portfolio diversification: evidence from emerging
financial markets, in Investing in Emerging Markets , Euromoney Books, 19976. Gibbons, Michael, Stephen A. Ross, and Jay Shanken, 1989, A test of the efficiency
of a given portfolio, Econometrica, V 57, Issue 5, pp.1121-1152.7. Hansen, L.P. and R. Jaganathan, 1991, Implications of Securities Market Data for
Models of Dynamic Economies, Journal of Political Economy , vol. 99 - 262.
8. Jorion, P. and D. Miller, 1997, Investing in emerging markets using depositoryreceipts, Emerging Markets Quarterly, Vol. 1, 7-13.
9. Kim, M., A. C. Szakmary, and I. Mathur, 2000, Price transmission dynamicsbetween ADRs and their underlying foreign securities, Journal of Banking and Finance, Vol. 24, 1359-1382.
10. Lins, Karl, and Deon Strickland, 2000, Do non-U.S. Firms Issue Equity on U.S.Stock Exchanges to Relax Capital Constraints?, NBER Working Paper .
11. Maroney, N, and A. Protopapadakis, 1999, The Book-to-market and Size Effects ina General Asset Pricing Model: Evidence from Seven National Markets, WorkingPaper , University of New Orleans.
12. Officer, D. T. and J. R. Hoffmeister, 1987, ADRs: a substitute for the real thing?, Journal of Portfolio Management, Vol 13, 61-65.
13. Solnik, International Finance, 2nd edition, Addision-Wessley, 199114. Wahab, M., and A. Khandwala, 1993, Why not diversify internationally with
ADRs?, Journal of Portfolio Management, Vol. 19, 75-82.15. Wahab, M., M. Lashgari, and R. Cohn, 1992, Arbitrage opportunity in the American
depository receipts market revisited, Journal of International Financial Markets Institutions, and Money, Vol 2, 97-130.
-
8/8/2019 05 International Diversification With ADRs
25/34
Table 1Characteristics of American Depository Receipts (ADRs)
Level I Level II Level III 144aDescription Unlisted Listed on major U.S.
exchangesOffered and listed onmajor U.S. exchanges
Private U.S. placement toqualified institutionalbuyers (QIBs)
Primary Exchange OTC pink sheets NYSE, AMEX, orNASDAQ
NYSE, AMEX, orNASDAQ
U.S. private placementmarket using PORTAL
SEC registration Registration StatementForm F-6
Registration StatementForm F-6
Form F-1 and F-6 forIPOs, full registration
Exempt
Accounting standard Home country standards.No GAAP reconciliationrequired
Only partialreconciliation forfinancials
Full GAAPreconciliation forfinancials
Home country standards.No GAAP reconciliationrequired
U.S. ReportingRequirements
Exemption under Rule12g3-2(b)
Form 20-F filedannually
Form F-20 filedannually; short forms F-2 and F-3 used only forsubsequent offerings
Private placement, as Rule144a or, new issue, asLevel III.
Capital Raising No capital raising.
Existing share only
No capital raising.
Existing share only
Capital raising with new
share issues
Capital raising with new
share issuesType of Offering Public Public Public PrivateTime to completion 9 weeks 14 weeks 14 weeks 7 weeksCosts to Enter Market $25,000 $200,000 700,000 $500,000 - $2,000,000 $250,000 - $500,000Source: The ADR Reference Guide, JP Morgan, 2000.
-
8/8/2019 05 International Diversification With ADRs
26/34
TABLE 2Number of ADRs listed Each Year from Each Country
Region Country uptoY80
Y81 Y82 Y83 Y84 Y85 Y86 Y87 Y88 Y89 Y80-Y90
Y90 Y91 Y92 Y93 Y94 Y95 Y91-Y95
Y96 Y97 Y98 Y99 Y00 Y96-Y01
Total
Australia 1 4 1 1 1 2 1 11 1 1 1 1 4 2 1 2 5 20Denmark 1 1 1 1 1 1 3France 0 1 1 2 1 1 6 7 5 3 2 6 23 29Germany 0 0 5 1 3 3 9 21 21Ireland 1 1 1 3 1 1 2 4 1 2 1 1 3 8 15Italy 1 2 3 1 2 2 1 6 2 2 4 13Japan 22 1 1 1 25 1 1 1 1 2 2 6 31Netherlands 3 1 1 1 6 1 2 1 4 1 4 2 3 3 13 23New Zealand 0 1 1 2 4 4 6Norway 0 1 1 2 2 2 4Spain 2 2 1 5 0 1 1 2 7Sweden 1 1 1 1 4 0 4 2 2 1 9 13Switzerland 0 0 2 1 2 5 10 10
DevelopedCountries
UK 9 2 3 4 5 3 3 29 2 1 3 6 2 5 19 8 10 9 13 15 55 103
Argentina 0 8 2 1 11 11Brazil 0 1 5 2 2 5 15 15Chile 0 1 1 4 8 2 16 17 7 1 1 9 25Columbia 0 1 1 1Mexico 0 2 2 3 8 7 22 3 1 2 2 1 9 33
LatinAmerica
PeruVenezuela
00
31 1 1
33
33
China 0 5 3 1 4 13 13Indonesia 0 3 3 3Korea 0 3 2 2 7 7Philippines 0 1 1 2 2Singapore 0 1 1 1 3 3Taiwan 0 1 1 3 5 5
Asia
Total 90 86 245 423
-
8/8/2019 05 International Diversification With ADRs
27/34
-
8/8/2019 05 International Diversification With ADRs
28/34
TABLE 4
Own Correlation Between ADR portfolio Returns and Country Indices
Developed CountriesPeriodAus Jap Den Neth Swe UK Fra Ita NZ Nor Spa Ire Ger Swi
1981-1990 0.50 0.80 0.45 0.85 0.74 0.93 - 0.59 - - 0.83 0.59 - -1991-1995 0.62 0.92 0.70 0.89 0.92 0.80 0.75 0.58 0.70 0.78 0.95 0.95 - -1996-2001 0.70 0.81 0.73 0.88 0.90 0.90 0.75 0.82 0.86 0.91 0.83 0.94 0.39 0.42
Latin AmericaPeriodChi Mex Col Per Ven Bra Arg
1981-1990 - - - - - - -1991-1995 0.36 0.34 - - - - -1996-2001 0.91 0.87 0.49 0.74 0.60 0.85 0.95
AsiaPeriodInd Kor Phi Sin Tai Chi
1981-1990 - - - - - -
1991-1995 - - - - - -1996-2001 0.70 0.69 0.68 0.56 0.65 0.91
-
8/8/2019 05 International Diversification With ADRs
29/34
Table 5Test Results of Index Model
The index model with risk-free rate is defined as, , , , ,c i i a i a i m i m i i
r r r = + + + 1,.....,i N = where, r c,i , r a,i , and r m,i are the excess returnsover risk free rate of each country indices, ADR portfolio returns, and U.S. market portfolio returns. All country indices and U.S. market index havebeen collected from MSCI. ADR returns have been collected from CRSP. We construct equally weighted ADR portfolio for each country. We have 6,12, and 15 countries from developed region for the period 1981-1990, 1991-1995, and 1996-2001 respectively. A total of 2 countries have been includedfor the period 1991-1995, and 7 countries for the period 1006-2001 in Latin American region. We have 6 Asian countries in the period 1996-2001. As aresult we have system of equations for each sample period. We hypothesize: 0 H : 0i = i . The rejection of null implies that ADR portfolio cannotprovide as much diversification benefits as the country indices. In panel A, we report the p-values of three different tests: Wald (chi-square), LikelihoodRatio, LR (F-statistics), and Gibbons, Ross, Shanken, GRS (F-statistics). In Panel 2, we test the hypothesis: 0 H : i = i , that is, each country
have the same alpha. In other words, the reject of the null implies that each country has different zero-beta asset.
Index ModelH 0: i =0 i
Index Model (zero beta)H 0: i= i
Wald Test LR Test GRS Test Wald Test LR Test
Time Period Region
P-value P-value P-value P-value P-value
1981-1990 Developed 0.000 0.000 0.000 0.004 0.021
Developed 0.000 0.000 0.000 0.003 0.0741991-1995
Latin America 0.296 0.306 0.323 0.816 0.871
Developed 0.154 0.052 0.276 0.125 0.107
Latin America 0.000 0.000 0.201 0.663 0.0581996-2001
Asia 0.000 0.000 0.000 0.297 0.401
-
8/8/2019 05 International Diversification With ADRs
30/34
TABLE 6Spanning Tests of Overidentifying Conditions: By Region
Rt is the n-dimensional vector of n-assets, and r t = R t E(R t) is the vector of excess returns. There are p benchmark assets, and the remaining, q =n-p aretest assets. The SDF mc, cj is the predetermined risk-free rate. We use two risk-free rate for the tests. If the n assets are spanned by the benchmark assets, then the volatility bound constructed from the benchmark and test assets remain unchanged when test assets are excluded. The volatility bound isformed by constructing two discount factors that have two expected values c1, and c2. The system that define the volatility bound for n assets is:
( )1 1,| t t t c R E R m v= = and ( )2 2,| t t t c R E R m v= = where [ ] [ ] [ ] E R Cov Rm E m = , and , , , ,,1 1
j l j l t k j k t j
p q
c t
l k
m r r c = =
= + + with ,( ) , 1, 2 jc t j E m c j= = and( )t t t r R E R= . The null hypothesis is that p benchmark assets are sufficient to span all n=p+q assets, It means that if q test assets are excluded from
the SDF, then the volatility bound does not change. In other words, the null hypothesis is: , 1 , 2 0l l = = . That is the overidentifying condition. For q test assets the number of overidentifying conditions are 2q . Hansen J-statistics are used to evaluate the conditions, which is a chi-square distributionwith degrees of freedom equal to 2q . GMM technique is used to estimate the equations. Our benchmark assets are all country indices, all ADRportfolios, and the U.S. market except the country indices (ADR portfolios) for a region.Panel A: Spanning tests: P-values
1981-1990 1991-1995 1996-2001P-values P- values P-values
Regions ExcludedIndex
Asymptotic Bootstrap Asymptotic Bootstrap Asymptotic BootstrapCountry 0.000 0.000 0.219 0.151 0.334 0.341Developed
ADR 0.000 0.001 0.199 0.108 0.302 0.308Country - - 0.000 0.003 0.016 0.009Latin America
ADR - - 0.000 0.001 0.013 0.022Country - - - - 0.070 0.102Asia
ADR - - - - 0.004 0.000
Panel B: Implication of spanning test results in panel ACan country plus U.S. marketindex mimic ADR returns?
Regions
Implications 1981-1990 1991-1995 1996-2001
Yes Yes ADR and country substitutes - DC DCYes No ADR required,
country not required - - AS No Yes Country required,
ADR not required - - -
Can ADRs plusU.S. market indexmimic countryreturns?
No No Both ADR and country required DC LA LA
-
8/8/2019 05 International Diversification With ADRs
31/34
TABLE 7Spanning Tests For Portfolios Sorted By Number of ADRs
In panel A, for each t-period returns, we form equally-weighted portfolios based on ( t-1)- period ranking of countries with respect to the number of ADRs for each country. We divide the countries in three different groups: largest, medium, and smallest based on the rank each year. We continue tofollow the same process for each years starting form 1980 to 2001. Then we stacked the data for the different periods, and conduct our spanning test.The benchmark assets consist of all three largest, medium, and smallest - ADR portfolios, three equally-weighted country portfolios, and the U.S.market except one of the ADR (country) portfolios that is considered as test asset. In panel B, we divide country and ADR portflios into the largest,medium, and the smallest groups based on the number of ADRs in the year 1990, 1995, and 2001. Hansen J-statistics are used to evaluate theoveridentifying conditions of the spanning test, which is a chi-square distribution with degrees of freedom equal to 2q . GMM technique is used toestimate the equations.Panel A:
1981-1990 1991-1995 1996-2001 1981-2001Portfolios ExcludedIndex P-values P-values P-values P-values
Large Country 0.000 0.000 0.299 0.000ADR 0.000 0.000 0.216 0.000
Medium Country 0.008 0.028 0.034 0.002ADR 0.001 0.001 0.009 0.000
Small Country 0.000 0.000 0.049 0.009ADR 0.018 0.000 0.046 0.000
Panel B:1981-1990 1991-1995 1996-2001Portfolios Excluded
Index P-values P-values P-valuesLarge Country 0.000 0.000 0.013
ADR 0.001 0.019 0.141Medium Country 0.000 0.020 0.024
ADR 0.003 0.003 0.019Small Country 0.000 0.000 0.027
ADR 0.000 0.015 0.007
-
8/8/2019 05 International Diversification With ADRs
32/34
TABLE 8Spanning Tests: By Country
For marginal diversification benefits, each time we exclude one country ADR portfolio (one country index), find theprobability of rejection of the null hypothesis that returns of all countries portfolios, ADRs portfolios and U.S.market can mimic that countrys ADR returns (returns of ADRs portfolios, all other countries portfolios and U.S.market can mimic that countrys returns) from corresponding J-statistics. Then we replace that countrys ADRportfolio (country index), and exclude another countrys ADR portfolio (country index), and test the null hypothesis.We repeat the process for all countries ADR portfolios and country indices. Countries are organized in threecategories based on the ranking by the number of ADRs as of year 2001.Panel A:Portfolio Excluded Index: Country Excluded Index: ADR Country
1981-1990
1991-1995
1996-2001
1981-1990
1991-1995
1996-2001
UK 0.000 0.047 0.701 0.000 0.766 0.006Japan 0.000 0.037 0.013 0.008 0.061 0.711
Large France - 0.155 0.045 - 0.232 0.177Mexico - 0.000 0.549 - 0.000 0.012Chili - 0.106 0.000 - 0.001 0.007Australia 0.022 0.006 0.629 0.492 0.000 0.000Netherlands 0.000 0.189 0.483 0.023 0.478 0.000
Sweden 0.032 0.116 0.101 0.073 0.120 0.339Italy 0.002 0.086 0.000 0.120 0.197 0.076Medium Sweden 0.032 0.116 0.101 0.073 0.120 0.339
Italy 0.002 0.086 0.000 0.120 0.197 0.076Ireland 0.241 0.111 0.019 0.140 0.117 0.006Brazil - - 0.009 - - 0.012Argentina - - 0.061 - - 0.071China - - 0.009 - - 0.000Denmark 0.357 0.335 0.064 0.205 0.507 0.749Spain 0.320 0.209 0.688 0.004 0.002 0.062New Zealand - 0.071 0.801 - 0.008 0.488Norway - 0.756 0.081 - 0.661 0.004Germany - - 0.752 - - 0.998Switzerland - 0.821 - - 0.013Colombia - - 0.000 - - 0.004
Small Peru - - 0.013 - - 0.061Venezuela - - 0.947 - - 0.929
Bolivia - - 0.924 - - 0.052Indonesia - - 0.039 - - 0.581Korea - - 0.008 - - 0.041Philippines - - 0.517 - - 0.815Singapore - - 0.007 - - 0.000Taiwan - - 0.002 - - 0.000
-
8/8/2019 05 International Diversification With ADRs
33/34
Panel B: Implication for spanning results in panel A
1981-1990 1991-1995 1996-2001Can
countryplusU.S.marketindexmimicADRreturns?
Implications
DevelopedCountry LatinAmerica Asia DevelopedCountry LatinAmerica Asia DevelopedCountry LatinAmerica Asia
Yes Yes
ADR and countrysubstitutes
Denmark Ireland
- - Denmark NetherlandsSwedenUKFranceItalyNorwayIreland
- - SwedenNewZealandGermanySpainDenmark
VenezuelaArgentina
Philippines
Yes No
ADRrequired,
country not required
Spain - - NewZealandSpain
Chile - AustraliaNetherlandsUKNorwaySwitzerland
MexicoBolivia
No YesCountryrequired,
ADR not required
AustraliaSwedenItaly
- - Japan - - JapanFranceItaly
Peru Indonesia
CanADRsplusU.S.marketindexmimiccountryreturns?
No No ADR and countryrequired
JapanNetherlandsUK
- - Australia Mexico - Ireland BrazilChileColumbia
KoreaSingaporeTaiwan
China
-
8/8/2019 05 International Diversification With ADRs
34/34
TABLE 9Ranking of Countries by Number of ADRs Each Year
Region Country Y80 Y81 Y82 Y83 Y84 Y85 Y86 Y87 Y88 Y89 Y90 Y91 Y92 Y93 Y94 Y95 Y96 Y97 Y98 Y99 Y00Australia 3 2 2 2 2 2 2 2 2 2 1 1 1 1 2 2 1 2 1 2 2Denmark 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3France 3 2 2 2 3 1 1 1 1Germany 3 3 2 3 3Ireland 3 3 3 3 3 3 3 3 3 2 2 3 2 2 2 2 2 2Italy 3 3 3 3 3 3 3 2 2 2 2 2 2 2Japan 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1Netherlands 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2New Zealand 3 3 3 2 3 3 3 3Norway 3 3 3 3 3 3 3 3Spain 3 3 2 2 2 2 2 3 3 3 3 3 3 3Sweden 3 3 3 3 3 3 3 3 3 3 2 2 2 3 3 3 2 2 2 2 2Switzerland 3 3 3 3 3
Developed
Countries
UK 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1Argentina 2 2 2 2 2Brazil 3 3 2 2 2
Chile 3 3 3 2 3 1 1 1 1 1 1Columbia 3 3 3 3 3Mexico 3 2 2 1 1 1 1 1 1 1 1
LatinAmerica
PeruVenezuela
33
33
33
33
33
China 3 2 2 2 2Indonesia 3 3 3 3 3Korea 3 3 3 3 3Philippines 3 3 3 3 3Singapore 3 3 3 3 3AsiaTaiwan 3 3 3 3 3